Slide 1: Creator of the Associative Processing Unit (APU) for AI
Lee-Lean Shu, Chairman, President & CEO Creator of the Associative Processing Unit Doug Schirle, CFO (APU) for AI and a leading provider of high- Didier Lasserre, VP of Sales and Investor performance memory solutions Relations August 2026
Slide 2: Safe Harbor
Safe Harbor The statements contained in this presentation that are not purely historical are forward-looking statements within the meaning of Section 21E of the Securities Exchange Act of 1934, as amended, including statements regarding GSI Technology’s expectations, beliefs, intentions, or strat eg
Slide 3: GSI By the numbers
GSI By the numbers LTM Revenue (as of Q1 FY2027) Cash and cash equivalents1 Market Cap2 $25.1M $77.0M $258.3M ( NO DEBT) Insider Ownership Years with TSMC SRAM chips shipped to date 20% 30 140M Employees Worldwide Patents Granted Founded / IPO 126 158 Total 1995 / 2007 96 for APU 1. Includes cash an
Slide 4: From SRAM Leader to Edge AI Compute Platform
From SRAM Leader to Edge AI Compute Platform transition into Edge AI enabled by 30+ years of semiconductor expertise Leveraging Decades of Memory Innovation to Build the Optimal Edge AI Processor High-Performance Developer of APU Emerging Edge AI SRAM Leadership Compute-in-Memory (CIM) Platform Lead
Slide 5: Why Edge AI requires a Different Processor
Why Edge AI requires a Different Processor Cloud AI Processors • Deliver exceptional AI performance, but are designed for data centers, not edge devices • Require substantial power and cooling infrastructure • Depend on moving data to and from the cloud, creating latency, bandwidth, and security cha
Slide 6: CIM reduces data movement, uses less power
CIM reduces data movement, uses less power GPU, CPU (Von Neumann model) APU CIM • Massive data transfers • Limited data movement • Higher power consumption / more heat created • Lower power consumption / less heat generated DRAM (memory) Large Distributed High-Bandwidth Memory L2 Cache Bit-Level Par
Slide 7: Targeted Edge Applications for GSI APU Family
Targeted Edge Applications for GSI APU Family Not every AI workload belongs in the cloud ©2026 GSI Technology, Inc. 7
Slide 8: Gemini-II in the “sweet spot” for Edge AI
Gemini-II in the “sweet spot” for Edge AI Bake-Off demonstrated best Performance-Per-Watt Platform Time to First Token (TTFT) System Power Gemini-II APU < 2.5 seconds ~30W Snapdragon X Elite 12 seconds ~30W Jetson Thor ~ 3 seconds >100W Gemini-II delivers up to 300% better performance-per-watt with
Slide 9: Gemini-II POCs: Demonstrate Repeatable Edge AI Platform
Gemini-ii POCS: Demonstrate Repeatable Edge AI Platform Current Platform Validation in Ongoing POCs • Defense/Tactical Edge: ”Sentinel” perimeter-security drone • Smart Infrastructure: ”Smart Guardian Eye” smart-city deployment Adjacent Market Opportunities for Future POCs • Industrial inspection •
Slide 10: Gemini-II: Targeting Markets Where Performance-Per-Watt Matters Most
Gemini-II: Targeting Markets Where Performance- Per-Watt Matters Most Selected markets share the power, latency, and real-time decision-making requirements where Gemini-II is most differentiated Target Edge AI Markets Ongoing Program Validation • Smart Infrastructure • Taiwan Smart City POC • Autono
Slide 11: Plato: Extending Edge AI to Autonomous Systems
Plato: Extending Edge AI to Autonomous Systems Designed to bring multimodal AI awareness and autonomous decision-making to power- constrained edge devices Gemini-II Plato—Next Generation APU • 15 W–100 W power profile • Target 2 W–10 W power profile • Real-time multimodal AI • 1/4 the footprint of G
Slide 12: Plato: Enabling Context-Aware Autonomous Edge AI
Plato: Enabling Context-Aware Autonomous Edge AI Plato Camera + Sensors → Understand Situation → Decide Action Gemini-II Camera → Act Autonomously Situational Awareness → Detect Object/Person → Understand Event → Provide Insight Edge AI Camera Today → Detect Person Plato extends GSI's edge AI platfo
Slide 13: APU Roadmap
APU Roadmap Start with specialized edge AI deployments; expand into broader AI markets CY2026 Gemini-II CY2028 Future Deployments Plato Enhanced APUs Specialized Edge AI Launch Expanding Edge Targets Broader Edge LLM Deployments Highlights • “Sentinel” G2 Tech Drone POC • Higher DRAM bandwidth for V
Slide 14: Building a Scalable Edge AI Platform
Building a Scalable Edge AI Platform Leveraging reusable software, AI-assisted development tools, and system integrator partnerships to expand across multiple edge AI markets Current Commercialization Strategy Building a Repeatable Growth Engine • Focus on a small number of high-value POCs • AI-assi
Slide 15: Financial overview—Q1 FY2027
Financial overview—Q1 FY2027 Sufficient Funds to commercialize Gemini-II & Advance Plato design Q1 FY2027 includes higher Cash from financing activities in Stable SRAM revenue from Q1 FY’27 were $14.0M including R&D related to Plato chip continued demand from GSI’s $9.3M in ATM proceeds. Cash design
Slide 16: High-Margin SRAM Business Funds APU Development
High-Margin SRAM Business Funds APU Development SRAM revenue provides financial foundation to develop and commercialize APU platform High density, low latency, high bandwidth, low power consumption Industry leader with largest portfolio of high-performance memory products SigmaQuad and SigmaDDR core
Slide 17: Radiation-Hardened & Tolerant SRAM
Radiation-Hardened & Tolerant SRAM for Aerospace, Satellite & Defense Advantages: Space Semiconductor TAM • Chips designed to protect from solar, deep space radiation Radiation Hardened, Radiation Tolerant • Highest density, fastest speed, with “pin-to-pin” replacement • Completed testing and string
Slide 18: Investment Highlights
Investment Highlights Building a Scalable Edge AI Platform Company • Proven semiconductor manufacturing model capable of supporting future APU growth • Strong financial foundation with $77.0M cash and no debt • Gemini-II validation through active Smart City, Defense, and U.S. Army programs • Gemini-
Slide 19: GSI Technology
GSI Technology High Performance Components for Leading-Edge Technology GSITechnology.com / IR Contact: GSIT@HaydenIR.com
Slide 20: Appendix
Appendix ©2026 GSI Technology, Inc. 20
Slide 21: Accessing DoW SBIRs & Government funds
Accessing DOW SBIRs & Government funds Offsetting R&D expenses & Creating APU awareness within D OW Agencies GSI awarded over $6 million of SBIR* funds to date • $2 million Phase II with U.S. Army to develop & test ruggedized edge AI platform • POC with G2 Tech along with DoW and a foreign defense a
Slide 22: APU Platform & roadmap
APU Platform & roadmap Chip Launch Timing 2015–2026 CY 2026 CY2027 APU Roadmap Gemini-I (28 nm) Gemini-II (16 nm) Plato (12 nm) Edge AI, including: Edge Search Multimodal GenAI/LLMs DB Index Builds SAR “On-board” Computer Vision for Target Applications SAR Computer Vision • Drones Fast Vector Search
Slide 23: Experienced Management Team
Experienced Management Team Years of Years with Name Title Prior Companies Experience GSI Lee-Lean Shu Chairman and CEO 47 30 Sony, AMD Doug Schirle Chief Financial Officer 47 26 Cypress, Pericom Didier Lasserre VP Sales and IR 38 28 Cypress, Solectron Avidan Akerib VP of Associative Computing 45 10
Slide 24: Income Statement
Income Statement CONDENSED CONSOLIDATED STATEMENTS OF OPERATIONS (in thousands, except per share data) (Unaudited) ©2026 GSI Technology, Inc. 24
Slide 25: Summary Balance Sheet
Summary Balance Sheet CONDENSED CONSOLIDATED BALANCE SHEETS (in thousands) (Unaudited) ©2026 GSI Technology, Inc. 25